Incipient Fault Diagnosis Method For IGBT Drive Circuit Based On Deep Learning

IGBT is an important electronic component of power electronics and the core device of electric energy conversion.It is applied to the field of power electronics from tens of watts to tens of megawatts.At present,IGBT has been widely used in consumer appliances,industrial control,track traction,new energy power generation,HVDC transmission and other fields,so it is particularly important to improve the reliability and stability of IGBT.Starting from the incipient fault diagnosis of IGBT drive circuit,this paper evaluates the operation status of IGBT and the speed of fault diagnosis.Incipient fault diagnosis of IGBT drive circuit,determine the type of fault according to the changes of various important parameters,and form the data information as the basis of state judgment to achieve timely and accurate maintenance.At the same time,it can extend the service life of the system and reduce the maintenance cost of the equipment,so as to achieve economical maintenance,which is very important for improving the reliability and safety of the entire system.This paper presents an incipient fault diagnosis method of IGBT drive circuit based on deep learning.Firstly,the Monte Carlo method is used to extract the time-domain response signal of the circuit under test as sample data.However,the existing fault diagnosis methods are not ideal for incipient fault feature extraction.In this paper,the advantages of stacked auto-encoders in deep learning in extracting essential features of data are used to extract features from sample data.The learning rate of each AE in SAE is optimized by quantum particle swarm optimization.In addition,for the feature information extracted by SAE,a multi-class relevance vector machine is used to establish an incipient fault diagnosis model for the obtained features.Because the width of the kernel function in RVM has an important influence on the diagnosis effect,this paper uses quantum-behaved particle swarm optimization(QPSO)to optimize the above parameters.The experimental results show that the deep learning method can effectively extract the instinct features of the incipient faults of the IGBT drive circuit.On this basis,the incipient fault multi-class RVM of the IGBT drive circuit can achieve100% diagnostic accuracy.